A Robust Alternate Least Squares Algorithm for Approximate Joint Diagonalization in Overdetermined Blind Source Separation.
Shinya SaitoKunio OishiPublished in: GCCE (2021)
Keyphrases
- blind source separation
- mixing matrix
- least squares
- natural gradient
- variable step size
- negative matrix factorization
- learning algorithm
- independent component analysis
- input data
- non stationary
- eigenvalue decomposition
- k means
- particle swarm optimization
- probabilistic model
- information retrieval
- clustering method
- convergence rate
- source separation
- cost function